On Fri, Jun 21, 2019 at 10:29 PM Bruno Marchal <[email protected]> wrote:
> On 21 Jun 2019, at 08:49, Bruce Kellett <[email protected]> wrote: > > On Fri, Jun 21, 2019 at 4:26 PM 'Brent Meeker' via Everything List < > [email protected]> wrote: > > On 6/20/2019 11:11 PM, Bruce Kellett wrote: >> >> After all, repetitions of the relevant interactions are happening all the >>> time: and not just in our controlled experiments. How can there be such >>> things as objective probabilities in the MWI scenario? How can we use >>> experimental evidence to support theories when we do not know whether our >>> observer probabilities are representative or not? >>> >>> >>> The same as in any probabilistic theory. We repeat it so many times >>> that we have statistics that we can compare to the theoretical >>> distribution. The same way you would test your theory that a coin was fair. >>> >> >> In other words, MWI is experimentally disconfirmed. >> >> How so? In repeated experiments I'm aware of (and a lot of photons go >> thru Aspect's EPR experiments) the statistics are consistent with the >> theory. To disconfirm MWI you'd have to observe statistics far from the >> expected value, which is why Tegmark proposed his machine gun suicide >> experiment. >> > > If you observe statistics far from those expected under the Born Rule you > just assume that your calculation of the wave function is in error! > > If MWI is true, then you would expect that in at least some cases, the > Born Rule would be disconfirmed. There necessarily exists branches of the > wave function in which this is the case. How can you be sure that were are > not on such a branch? > > On some branches, you can send a large number of photons to your half > silvered mirror, and observe that the results conform to binomial > statistics with p = 0.5. But then next long sequence of photons will all go > just one way, casting doubt on your earlier statistics. Since such branches > necessarily exist under MWI, how can one ever have confidence in the > results of any quantum experiment? > > In other words, in order to do experiments in quantum optics, one has to > assume that MWI is false. > > > I don’t see this at all. In the iterated duplication experience, it is > true that there is a guy who saw only Moscow, but there are few chance it > belongs to the random sample that we use when verifying statistics, > You are using the 3p perspective to interpret a 1p probability -- you make the mistake that you often accuse others of making. > and if he does, well, he will feel alone, and not convince its peers. That > works for both Everett, digital mechanism, but also any applied probability > experience. We can always expect deviation. The point is that we should not > count on it. That is why, even if Everett is true, you will take the lift > and not jump out of the window. > But you do not jump out of the window because you believe that the laws of classical physics apply. This is not a quantum choice, so Everett is irrelevant. > That is why in the iterated self-duplication version, I ask question like > “what is the better bet, white noise of a the binary expansion of PI?”. > Then the verification is made on some sample, and most will confirms “white > noise”, and virtually none will confirm “binary expansion of PI”. > I am talking about experiments in quantum optics, not your toy duplication thought experiments. > Bruce, your answer “I don’t know the probabilities” is strictly speaking > correct … for all application of probability. It trivialise applied > probability theory, it seems to me. > It does not trivialise probability in the single world picture, in which theories give probabilities that can be confirmed or disconfirmed by experiments. It only shows that probabilities have no meaning in MWI where all outcomes are certain to occur. In the classical duplication situation, the 1p understanding of probability is closer to how probability is used in science than the 3p notion that you are seeking to impose. > In the n-iteration with n low, all the copies can discuss together and see > that the binomial distribution is exactly verified. The number of copies > having seen W or M are exactly given by the triangle of Pascal. That > remains trivially true for large n, even if not explicitly testable, so > your “I don’t know the probabilities” becomes a reason to use the binomial > distribution. > But this is again the 3p picture. Which is not available in the quantum case. Bruce > Of we cannot know this for sure, if only because we cannot trust the > protocol, and maybe we were each time multiplied in 3, with a copy secretly > build in Vienna, perhaps. This shows that even the “verification” proves > nothing, but that is why in the theory we insist on the fact that the > protocol is verified, that mechanism is assumed etc, and derive theoretical > probabilities, and test them accordingly. > > Of course, the interesting scenario is given by the universal dovetailer, > where we are “multiplied” by infinity, and do the math to see if the > probabilities obeys to what we observe in nature or not. > > Bruno > -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/CAFxXSLT5eNNtYLTd71v%2BkXehv2jXhfeM2AiGBGiOAzmCPsq3ww%40mail.gmail.com.
